Convert `LEFT JOIN` to `INNER JOIN` when the result of `LEFT JOIN` can
never be different from the result of an `INNER JOIN`
This is useful because 1. it uses less instructions and 2. it allows for
join reordering due to inner join commutativity
Reviewed-by: Preston Thorpe <preston@turso.tech>
Closes#2972
current code is incorrectly assuming that `index_cursor.rowid()` always
finds a rowid, but this is not the case when `NullRow` has previously
been called.
Reviewed-by: Preston Thorpe <preston@turso.tech>
Closes#2973
#2928 was reverted, but let's at least add the ability to create an
equivalent big schema so that the issues reported by Preston using that
larger schema can be reproduced and fixed.
Reviewed-by: Preston Thorpe <preston@turso.tech>
Closes#2960
The MakeRecord instruction now accepts an optional affinity_str
parameter that applies column-specific type conversions before creating
records. When provided, the affinity string is applied
character-by-character to each register using the existing
apply_affinity_char() function, matching SQLite's behavior.
Fixes#2040Fixes#2041
only use `self.stack.top()` when we need to store `Arc<Page>` in struct
```sh
Execute `SELECT count() FROM users`/limbo_execute_select_count
time: [5.3733 µs 5.3801 µs 5.3881 µs]
change: [-34.047% -33.949% -33.851%] (p = 0.00 < 0.05)
Performance has improved.
Found 1 outliers among 100 measurements (1.00%)
1 (1.00%) high severe
```
Reviewed-by: Jussi Saurio <jussi.saurio@gmail.com>
Closes#2962
- Use github's default runner due blacksmith's higher performance
variance - Always comment on PRs about perf variances (currently only
for people with write permissions)
Closes#2949
This fixes a special case of the behavior described in #2501 - the
special case is that WHERE conditions that were selected as seek
predicates for the left join table were not properly evaluated when the
right-hand-side table returned no match.
The test in commit 12d72d115588a9e744bdb22382998ba1bf9031ab should
demonstrate this adequately - this should return no rows, but on `main`
it returns `1|NULL`.
Closes#2955
Closes#2946
currently we always evaluate the binary expression, then coerce it to
zero/null with the `ZeroOrNull` instruction, and then emit a separate
jump.
this is fine for non-conditional expressions where we are using the
value itself (e.g. in a SELECT result column), but in conditionals we
don't care about that at all and just want to either jump or not jump.
so: try to keep the spirit of code reuse, but still have distinct
implementations for conditionals and non-conditionals.
Reviewed-by: Preston Thorpe <preston@turso.tech>
Closes#2947
Attempt to improve performance of `SELECT COUNT(*) FROM t WHERE 1 = 1`
(redundant filter added to force sqlite and tursodb to iterate over rows
instead of applying count optimization).
Before that PR turso with hot cache runs for ~440ms.
After that PR turso with hot cache runs for ~200ms.
This PR eliminates small things on a hot path and also introduce
separate flow for `get_next_record` which just increment cell idx if its
safe to do that - without much additional overhead.
Also, this PR eliminates RefCell/Cell from PageStack because there were
completely unnecessary.
Reviewed-by: Preston Thorpe <preston@turso.tech>
Reviewed-by: Pere Diaz Bou <pere-altea@homail.com>
Closes#2887
